\documentclass[12pt]{article}
\usepackage[round]{natbib} \setlength{\bibsep}{2pt}
\usepackage{amsmath}
%\usepackage{fullpage}
\usepackage{pslatex}
\usepackage{longtable}
%\usepackage{appendix}
\usepackage{url}
\usepackage{setspace}
\usepackage{sectsty}
\usepackage[left=1in,top=1in,right=1in,bottom=1in]{geometry}
%\usepackage{lscape}
\usepackage{pdflscape}
\usepackage{threeparttable}
%\usepackage{color}
\usepackage{rotating}
\usepackage{caption}
%\usepackage[pdftex]{graphicx}
\usepackage{epsfig}
\usepackage{float}
\usepackage{subfig}
\usepackage[section]{placeins} %To prevent floats from floating past their section

% To restart numbering of tables/figures in each section 
\usepackage{chngcntr}
\counterwithin{figure}{section}
\counterwithin{table}{section}

% To have the sections labels alphabetically and tables accordingly.
\renewcommand*{\thesection}{\Alph{section}}
\renewcommand\thetable{\Alph{section}.\arabic{table}}
\renewcommand\thefigure{\Alph{section}.\arabic{figure}}

\renewcommand{\baselinestretch}{2}

\urldef{\arunacha}\url{arunacha@umich.edu}
\urldef{\watson}\url{watson.584@osu.edu}
%\newcommand{\argmax}{\operatornamewithlimits{argmax}}

%\newcommand{\bibfont}{\small}
\allsectionsfont{\normalsize\bf}

%\setlongtables
%\makeatletter
%\def\table{\@ifnextchar[{\table@i}{\table@i[\fps@table]}}
%\def\table@i[#1]{\@float{table}[#1]\tiny}
%\makeatother

% Alter some LaTeX defaults for better treatment of figures:
    % See p.105 of "TeX Unbound" for suggested values.
    % See pp. 199-200 of Lamport's "LaTeX" book for details.
    %   General parameters, for ALL pages:
    \renewcommand{\topfraction}{0.9}	% max fraction of floats at top
    \renewcommand{\bottomfraction}{0.8}	% max fraction of floats at bottom
    %   Parameters for TEXT pages (not float pages):
    \setcounter{topnumber}{2}
    \setcounter{bottomnumber}{2}
    \setcounter{totalnumber}{4}     % 2 may work better
    \setcounter{dbltopnumber}{2}    % for 2-column pages
    \renewcommand{\dbltopfraction}{0.9}	% fit big float above 2-col. text
    \renewcommand{\textfraction}{0.07}	% allow minimal text w. figs
    %   Parameters for FLOAT pages (not text pages):
    \renewcommand{\floatpagefraction}{0.7}	% require fuller float pages
	% N.B.: floatpagefraction MUST be less than topfraction !!
    \renewcommand{\dblfloatpagefraction}{0.7}	% require fuller float pages
	% remember to use [htp] or [htpb] for placement


%%%% To keep figures on separate pages
\makeatletter
\@fpsep\textheight
\makeatother
%%%%



\begin{document}

\newenvironment{margins}[2]
{
\begin{list}{} {
\setlength{\leftmargin}{#1}
\setlength{\rightmargin}{#2}
} \item }
{\end{list}}
%\bibliographystyle{ier}%\bibliographystyle{econometrica}%\bibliographystyle{apsr}
\bibliographystyle{apsrRAJ}
\renewcommand{\refname}{Appendix: References}

% Define macros

\newcommand{\defcvote}{``supports Conservative party''}
\newcommand{\defturnout}{``Turnout for 2005 General Election''}
\newcommand{\defconvote}{``Voted for Conservative Party in 2005 General Election''}
\newcommand{\defgjd}{``It is the government's responsibility to provide a job for everyone who wants one''}
\newcommand{\defped}{``Private enterprise is the best way to solve the UK's economic problems''}
\newcommand{\defpsd}{``Major public services and industries ought to be in state ownership''}
\newcommand{\defuld}{``The government should place an upper limit on the amount of money that any one person can make''}
\newcommand{\defgjo}{``It is the government's responsibility to provide a job for everyone who wants one'', from -2 (strongly disagree) to 2 (strongly agree)}
\newcommand{\defpeo}{``Private enterprise is the best way to solve the UK's economic problems'', from -2 (strongly disagree) to 2 (strongly agree)}
\newcommand{\defpso}{``Major public services and industries ought to be in state ownership'', from -2 (strongly disagree) to 2 (strongly agree)}
\newcommand{\defulo}{``The government should place an upper limit on the amount of money that any one person can make'', from -2 (strongly disagree) to 2 (strongly agree)}
\newcommand{\defincome}{``Real Income ('000s of pounds)''}
\newcommand{\statsig}{Statistical significance: $^{*}$10\% ; $^{**}$5\% ; $^{***}$1\%}
\newcommand{\secondstage}{Second stage of two-stage least squares regression}
\newcommand{\firststage}{First stage of two-stage least squares regression}
\newcommand{\leastsquares}{Least squares regression}

\newcommand{\tablespecs}{Columns (1)-(5) report results for whole sample.  Column (3) includes a measure of cognitive ability and (4)-(5) include family background measures as controls.  Columns (6)-(7) restrict the sample to prime-age (age 33-65); (8)-(9) expand the sample to include Northern Ireland; (10)-(11) restrict to women; and (12)-(13) restrict to men.  Omitted religion variable is `no religion'.  All specifications include region dummies and heteroskedasticity-robust standard errors clustered by household.  \statsig.}
\newcommand{\tablespecsturnout}{Columns (1)-(5) report results for whole sample.  Column (3) includes a measure of cognitive ability and (4)-(5) include family background measures as controls.  Columns (6)-(7) restrict the sample to prime-age (age 33-65); (8)-(9) expand the sample to include Northern Ireland; (10)-(11) restrict to women; and (12)-(13) restrict to men.  Column (14) adds free time as a control.  Omitted religion variable is `no religion'.  All specifications include region dummies and heteroskedasticity-robust standard errors clustered by household.  \statsig.}





\title{Height, Income and Voting\\ Web Appendix}
\author{}
\date{January 2015}

\maketitle\thispagestyle{empty}
\newcommand\comment[1]{}

\pagebreak 

\singlespacing

\section{Table of Contents}
\listoffigures
\listoftables

\newpage


\section{Variables and Summary Statistics}
%\renewcommand{\baselinestretch}{1}

\begin{footnotesize} 
\begin{center}
\begin{longtable}{p{4.3cm}p{1.5cm}p{9.5cm}}
%\input{table_variables.tex}
\caption{Variable Descriptions}
\end{longtable}
\end{center}
\end{footnotesize}

\begin{table}[H]
\begin{scriptsize}
\begin{center}
\input{summarystats16.tex}\label{tab:summarystats}
\caption{Summary Statistics}
\end{center}
Note: Survey-weighted using cross-sectional weights (excluding Northern Ireland) provided by the BHPS.  Unless explicitly noted, here and throughout this Appendix, data is from Wave 16 of the BHPS.
\end{scriptsize}
\end{table}

\begin{figure}[H]
  \begin{center}
  \subfloat[Height]{\includegraphics[scale=0.75]{histogram_height16.eps}\label{fig:histogram16}} \hfill
  \subfloat[Income]{\includegraphics[scale=0.75]{income_dist16.eps}\label{fig:incomedist16}}
  \caption{Distributions of Income and Height}
	\end{center}
\begin{footnotesize}
Note: \subref{fig:histogram16} Histogram of height by gender with normal distributions overlaid; \subref{fig:incomedist16} kernel density of real income by gender. 
\end{footnotesize}
\end{figure}

\newpage

\section{Reduced Form}

\begin{figure}[H]
  \begin{center}
	\subfloat[Support]{\includegraphics[scale=0.75]{fig_RF_MF_cvote_16.eps}\label{RFcvote_MF}}\\
  \subfloat[Vote]{\includegraphics[scale=0.75]{fig_RF_MF_convote_16.eps}\label{RFconvote_MF}}
  \caption{Reduced Form, Women vs. Men}
	\end{center}
\begin{footnotesize}

Note: Running line smooths of \subref{RFcvote_MF} \defcvote; and \subref{RFconvote_MF} \defconvote \ by gender, adjusted for age.  The X's mark the 10th and 90th percentile of the height distributions.  95 percent pointwise bootstrapped confidence intervals displayed.   
\end{footnotesize}
\end{figure}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{regRFcvote16.tex}
\caption{Reduced Form: Supports Conservative}
\end{center}
Note: Linear probability model; dependent variable is \defcvote. \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{logitRFcvote16.tex}
\caption{Reduced Form Logit: Supports Conservative}
\end{center}
Note: Logit model reporting average marginal effects; dependent variable is \defcvote.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{probitRFcvote16.tex}
\caption{Reduced Form Probit: Supports Conservative}
\end{center}
Note: Probit model reporting average marginal effects; dependent variable is \defcvote.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\clearpage


\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{regRFconvote16.tex}
\caption{Reduced Form: Vote Conservative}
\end{center}
Note: Linear probability model; dependent variable is \defconvote.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{logitRFconvote16.tex}
\caption{Reduced Form Logit: Vote Conservative}
\end{center}
Note: Logit model reporting average marginal effects; dependent variable is \defconvote.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{probitRFconvote16.tex}
\caption{Reduced Form Probit: Vote Conservative}
\end{center}
Note: Probit model reporting average marginal effects; dependent variable is \defconvote.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{regRFpeo14.tex}
\caption{Reduced Form: Private Enterprise Best}
\end{center}
Note: \leastsquares; dependent variable is \defpeo.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{ologitRFpeo14.tex}
\caption{Reduced Form Ordered Logit: Private Enterprise Best}
\end{center}
Note: Ordered logit model reporting average marginal effects; dependent variable is \defpeo.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{oprobitRFpeo14.tex}
\caption{Reduced Form Ordered Probit: Private Enterprise Best}
\end{center}
Note: Dependent variable is \defpeo.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{regRFpso14.tex}
\caption{Reduced Form: State Ownership}
\end{center}
Note: Dependent variable is \defpso.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{ologitRFpso14.tex}
\caption{Reduced Form Ordered Logit: State Ownership}
\end{center}
Note: Dependent variable is \defpso.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{oprobitRFpso14.tex}
\caption{Reduced Form Ordered Probit: State Ownership}
\end{center}
Note: Dependent variable is \defpso.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\clearpage

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{regRFgjo14.tex}
\caption{Reduced Form: Government Should Provide Jobs for All}
\end{center}
Note: Dependent variable is \defgjo.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{ologitRFgjo14.tex}
\caption{Reduced Form Ordered Logit: Government Should Provide Jobs for All}
\end{center}
Note: Dependent variable is \defgjo.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{oprobitRFgjo14.tex}
\caption{Reduced Form Ordered Probit: Government Should Provide Jobs for All}
\end{center}
Note: Dependent variable is \defgjo.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{regRFulo16.tex}
\caption{Reduced Form: Upper Limit on Income}
\end{center}
Note: Dependent variable is \defulo.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{ologitRFulo16.tex}
\caption{Reduced Form Ordered Logit: Upper Limit on Income}
\end{center}
Note: Dependent variable is \defulo.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{oprobitRFulo16.tex}
\caption{Reduced Form Ordered Probit: Upper Limit on Income}
\end{center}
Note: Dependent variable is \defulo.  \tablespecs
\end{scriptsize}
\end{sidewaystable}



\newpage

\section{Income and Height by Sex}

\begin{figure}[htp]
  \begin{center}
 \subfloat[Whole Sample]{\includegraphics[scale=0.75]{fig_IV1stAcvote16.eps}\label{IV1stA}} \hfill
 \subfloat[Women vs. Men]{\includegraphics[scale=0.75]{fig_IV1stMFcvote16.eps}\label{IV1stMF}}
  \caption{Income and Height}
	\end{center}
\begin{footnotesize}

Note: Running line smooths of real income, adjusted for age and gender (panel \subref*{IV1stA}) and age (panel \subref*{IV1stMF}).  The X's mark the 10th and 90th percentile of the height distributions.  95 percent pointwise bootstrapped confidence intervals displayed.   
\end{footnotesize}
\end{figure}


\newpage

\section{OLS, IV Second Stage, IV First Stage}

\begin{sidewaystable}
\begin{scriptsize}
\begin{center}
\input{regIV2cvote16.tex}
\caption{OLS: Supports Conservative}
\end{center}
Note: Linear probability model; dependent variable is \defcvote. \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}
\begin{scriptsize}
\begin{center}
\input{ivreg2IV2cvote16.tex}
\caption{IV Second Stage: Supports Conservative}
\end{center}
Note: Linear probability model; dependent variable is \defcvote.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}
\begin{scriptsize}
\begin{center}
\input{regIV1cvote16.tex}
\caption{IV First Stage: Supports Conservative}
\end{center}
Note: Linear probability model; dependent variable is \defincome.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}
\begin{footnotesize}
\begin{center}
\input{probitIV2cvote16.tex}
\caption{Probit: Supports Conservative}
\end{center}
Note: Probit model reporting average marginal effects; dependent variable is \defcvote.  \tablespecs
\end{footnotesize}
\end{sidewaystable}



\begin{sidewaystable}
\begin{scriptsize}
\begin{center}
\input{ivprobitIV2cvote16.tex}
\caption{IV Probit Second Stage: Supports Conservative}
\end{center}
Note: Instrumental variables probit model reporting average marginal effects; dependent variable is \defcvote.  \tablespecs
\end{scriptsize}
\end{sidewaystable}


\begin{sidewaystable}
\begin{scriptsize}
\begin{center}
\input{regIV2convote16.tex}
\caption{OLS: Vote Conservative}
\end{center}
Note: Linear probability model; dependent variable is \defconvote. \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}
\begin{scriptsize}
\begin{center}
\input{ivreg2IV2convote16.tex}
\caption{IV Second Stage: Vote Conservative}\label{IV2nd_convote16}
\end{center}
Note: Linear probability model; dependent variable is \defconvote.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}
\begin{footnotesize}
\begin{center}
\input{probitIV2convote16.tex}
\caption{Probit: Vote Conservative}
\end{center}
Note: Probit model reporting average marginal effects; dependent variable is \defconvote.  \tablespecs
\end{footnotesize}
\end{sidewaystable}

\begin{sidewaystable}
\begin{scriptsize}
\begin{center}
\input{ivprobitIV2convote16.tex}
\caption{IV Probit Second Stage: Vote Conservative}
\end{center}
Note: Instrumental variables probit model reporting average marginal effects; dependent variable is \defconvote.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}
\begin{scriptsize}
\begin{center}
\input{regIV2peo14.tex}
\caption{OLS: Private Enterprise Best}
\end{center}
Note: Dependent variable is \defpeo. \tablespecs
\end{scriptsize}
\end{sidewaystable}

\clearpage

\begin{sidewaystable}
\begin{scriptsize}
\begin{center}
\input{ivreg2IV2peo14.tex}
\caption{IV Second Stage: Private Enterprise Best}
\end{center}
Note: Dependent variable is \defpeo.  \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}
\begin{scriptsize}
\begin{center}
\input{regIV2pso14.tex}
\caption{OLS: State Ownership}
\end{center}
Note: Dependent variable is \defpso. \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}
\begin{tiny}
\begin{center}
\input{ivreg2IV2pso14.tex}
\caption{IV Second Stage: State Ownership}
\end{center}
Note: Dependent variable is \defpso.  \tablespecs
\end{tiny}
\end{sidewaystable}


\begin{sidewaystable}
\begin{scriptsize}
\begin{center}
\input{regIV2gjo14.tex}
\caption{OLS: Government Should Provide Jobs for All}
\end{center}
Note: Dependent variable is \defgjo. \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}
\begin{tiny}
\begin{center}
\input{ivreg2IV2gjo14.tex}
\caption{IV Second Stage: Government Should Provide Jobs for All}
\end{center}
Note: Dependent variable is \defgjo.  \tablespecs
\end{tiny}
\end{sidewaystable}


\begin{sidewaystable}
\begin{scriptsize}
\begin{center}
\input{regIV2ulo16.tex}
\caption{OLS: Upper Limit on Income}
\end{center}
Note: Dependent variable is \defulo. \tablespecs
\end{scriptsize}
\end{sidewaystable}

\begin{sidewaystable}
\begin{tiny}
\begin{center}
\input{ivreg2IV2ulo16.tex}
\caption{IV Second Stage: Upper Limit on Income}
\end{center}
Note: Dependent variable is \defulo.  \tablespecs
\end{tiny}
\end{sidewaystable}

\clearpage


\begin{table}
\begin{footnotesize}
\begin{center}
\input{youth16.tex}
\caption{Effect of Parents' Characteristics on Childrens' Height}
\end{center}
Note: Least squares regression; dependent variable is height (inches).  Father's and mother's support is a dummy indicating \defcvote. Sample includes children age 11-17.  Omitted religion variable is `no religion'.  All specifications include region dummies and heteroskedasticity-robust standard errors clustered by household.  \statsig.
\end{footnotesize}
\end{table}

\clearpage
\newpage

\section{Fixed Effects Models}

\begin{table}[htp]
\begin{footnotesize}
\begin{center}
\input{xtlogitwincome.tex}
\caption{Fixed Effects Logit}\label{xtlogit_wincome}
\end{center}
Note: Logit model with individual fixed effects;  dependent variable is \defcvote.  \statsig.
\end{footnotesize}
\end{table}

\newpage

\section{Over Time}

\begin{figure}[H]
%	\begin{centering}	
  \subfloat[Basic]{\includegraphics[scale=0.4]{OLSIV2nd_whole1.eps}\label{fig:OLSIV2nd_whole1}} \quad
  \subfloat[Extended]{\includegraphics[scale=0.4]{OLSIV2nd_whole2.eps}\label{fig:OLSIV2nd_whole2}}\\
  \subfloat[Ext + Cognition]{\includegraphics[scale=0.3]{OLSIV2nd_whole3.eps}\label{fig:OLSIV2nd_whole3}} \quad
  \subfloat[Ext + Parents Schooling]{\includegraphics[scale=0.3]{OLSIV2nd_whole4.eps}\label{fig:OLSIV2nd_whole4}} \quad
  \subfloat[Ext + HGS]{\includegraphics[scale=0.3]{OLSIV2nd_whole5.eps}\label{fig:OLSIV2nd_whole5}} \\
  \subfloat[Prime Age- Basic]{\includegraphics[scale=0.4]{OLSIV2nd_whole6.eps}\label{fig:OLSIV2nd_whole6}} \quad
  \subfloat[Prime Age- Extended]{\includegraphics[scale=0.4]{OLSIV2nd_whole7.eps}\label{fig:OLSIV2nd_whole7}}\\
  \subfloat[Incl. NI- Basic]{\includegraphics[scale=0.4]{OLSIV2nd_whole8.eps}\label{fig:OLSIV2nd_whole8}} \quad
	\subfloat[Incl. NI - Extended]{\includegraphics[scale=0.4]{OLSIV2nd_whole9.eps}\label{fig:OLSIV2nd_whole9}}\\
    \begin{footnotesize}
	\caption{Income Gradient Over Time: OLS vs IV}
	\label{fig:incomegradientovertimeW}
  \end{footnotesize}
	\begin{scriptsize}	
Note: Dependent variable is \defcvote.     
\end{scriptsize}
%\end{centering}
\end{figure}

\comment{
\begin{figure}[H]
%	\begin{centering}	
  \subfloat[Basic]{\includegraphics[scale=0.4]{OLSIV2nd_MF1.eps}\label{fig:OLSIV2nd_MF1}} \quad
  \subfloat[Extended]{\includegraphics[scale=0.4]{OLSIV2nd_MF2.eps}\label{fig:OLSIV2nd_MF2}}\\
  \subfloat[Ext + Cognition]{\includegraphics[scale=0.3]{OLSIV2nd_MF3.eps}\label{fig:OLSIV2nd_MF3}} \quad
  \subfloat[Ext + Parents Schooling]{\includegraphics[scale=0.3]{OLSIV2nd_MF4.eps}\label{fig:OLSIV2nd_MF4}} \quad
  \subfloat[Ext + HGS]{\includegraphics[scale=0.3]{OLSIV2nd_MF5.eps}\label{fig:OLSIV2nd_MF5}} \\
  \subfloat[Prime Age- Basic]{\includegraphics[scale=0.4]{OLSIV2nd_MF6.eps}\label{fig:OLSIV2nd_MF6}} \quad
  \subfloat[Prime Age- Extended]{\includegraphics[scale=0.4]{OLSIV2nd_MF7.eps}\label{fig:OLSIV2nd_MF7}}\\
  \subfloat[Incl. NI- Basic]{\includegraphics[scale=0.4]{OLSIV2nd_MF8.eps}\label{fig:OLSIV2nd_MF8}} \quad
	\subfloat[Incl. NI - Extended]{\includegraphics[scale=0.4]{OLSIV2nd_MF9.eps}\label{fig:OLSIV2nd_MF9}}\\
    \begin{footnotesize}
	\caption{Income Gradient Over Time: Women vs. Men}
	\label{fig:}
  \end{footnotesize}
	\begin{scriptsize}	
\end{scriptsize}
%\end{centering}
\end{figure}



\begin{table}
\begin{footnotesize}
\begin{center}
\input{xtregtotal_contrib052113.tex}
\caption{Fixed Effects: Total Contributory}\label{xtreg_total_contrib}
\end{center}
Note: Linear probability model with individual fixed effects;  dependent variable is \defcvote.  
\end{footnotesize}
\end{table}

\begin{table}
\begin{footnotesize}
\begin{center}
\input{xtlogittotal_contrib052113.tex}
\caption{Fixed Effects Logit: Total Contributory}\label{xtlogit_total_contrib}
\end{center}
Note: Logit model with individual fixed effects;  dependent variable is \defcvote.  Heteroskedasticity-robust standard errors clustered by individual.  
\end{footnotesize}
\end{table}

\begin{table}
\begin{footnotesize}
\begin{center}
\input{xtregtotal_universal052113.tex}
\caption{Fixed Effects: Universal}\label{xtreg_total_universal}
\end{center}
Note: Linear probability model with individual fixed effects;  dependent variable is \defcvote.  
\end{footnotesize}
\end{table}

\begin{table}
\begin{footnotesize}
\begin{center}
\input{xtlogittotal_universal052113.tex}
\caption{Fixed Effects Logit: Universal}\label{xtlogit_total_universal}
\end{center}
Note: Logit model with individual fixed effects;  dependent variable is \defcvote.  
\end{footnotesize}
\end{table}

\begin{table}
\begin{footnotesize}
\begin{center}
\input{xtregtotal_meanstest052113.tex}
\caption{Fixed Effects: Means-Tested}\label{xtreg_total_meanstest}
\end{center}
Note: Linear probability model with individual fixed effects and year dummies;  dependent variable is \defcvote.  
\end{footnotesize}
\end{table}

\begin{table}
\begin{footnotesize}
\begin{center}
\input{xtlogittotal_meanstest052113.tex}
\caption{Fixed Effects Logit: Means-Tested}\label{xtlogit_total_meanstest}
\end{center}
Note: Logit model with individual fixed effects and year dummies;  dependent variable is \defcvote.  
\end{footnotesize}
\end{table}

}

\clearpage
\newpage
\section{Alternative Channels: Height, Health and Risk Aversion}
Two questions not discussed in detail in the paper are whether height may shape political preferences in part through health or through risk tolerance, which would contaminate our interpretation of the height effect on preferences.  We discuss each of these in turn here.

\subsection{Height and Health}
With respect to health, the central concern is that if height is fundamentally a proxy for good health---and if healthy people need the National Health Service less---then they may be less inclined to defend parties that prioritize the NHS.  If so, height operates through both income and health, contaminating our interpretation of the height effect and violating the exclusion restriction in the IV.
 
We believe this particular channel is less likely to be a concern in our setting.  The literature generally finds that lower income individuals actually utilize the NHS no more (and, in several studies, less) than their affluent counterparts (Evandrou et al 1992; Cooper et al 1998; Goddard and Smith, 2001; Morris et al, 2006; for an extensive review see Dixon et al., 2007).  Insofar as the income gradient of NHS utilization is negative, and if higher utilization indeed directly predicts left-leaning political preference, this would mean that controlling for utilization should generate stronger effects of height on political outcomes in the reduced form and income on political outcomes in the IV.


\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{height_RF_health.tex}
	\label{tab:RF_health}
\caption{Reduced Form: Height and Health Utilization }
\end{center}
Note: Dependent variable is \defcvote.  Least squares regression of ``supports Conservative party'' on height, controlling for ``average annual number of inpatient days in NHS hospitals'' in columns (1)-(2) and ``average annual number of NHS services used'' in columns (3)-(4). Conditional models additionally control for health status.  All specifications include region dummies and heteroskedasticity-robust standard errors clustered by household.  \statsig.
\end{scriptsize}
\end{sidewaystable}


\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{height_IV2_health.tex}
	\label{tab:IV_health}
\caption{IV Second Stage: Height and Health Utilization }
\end{center}
Note: Dependent variable is \defcvote.  Two-stage least squares regression of ``supports Conservative party'' on height, controlling for ``average annual number of inpatient days in NHS hospitals'' in columns (1)-(2) and ``average annual number of NHS services used'' in columns (3)-(4). Conditional models additionally control for health status.  All specifications include region dummies and heteroskedasticity-robust standard errors clustered by household.  \statsig.
\end{scriptsize}
\end{sidewaystable}

Fortunately the BHPS allows us to address this issue head-on as it contains questions about NHS utilization.  First, we incorporate these health utilization measures as a control in the reduced form and instrumental variable regressions (see column (1) and (3) of Appendix Tables H.1 and H.2).  Here we see that the point estimate on height and on income instrumented with height remain nearly unaltered from those reported in Tables 1 and 5 of the paper.  As we would expect, the inclusion of these health variables did not alter the point estimates on height (nor those on income instrumented with height).    

However, these models, which follow the aforementioned studies of NHS utilization by including a measure of health problems to capture need for health services, arguably do not get at the stated concern directly. This is because height may predict utilization when not controlling for need.  In other words, the reviewer’s comment is really about the unconditional relationship between height and NHS utilization.  Therefore, we re-estimate the models described above without controlling for health problems.  In these unconditional models, reported in columns (2) and (4) of Appendix Tables H.1 and H.2, the point estimate and statistical significance does not change.  


\subsection{Height and Risk Tolerance}
We also investigate the possibility that height might affect be related to risk-seeking behavior and that risk-tolerant individuals may be less likely to support redistributive government programs; as we noted above, if this were true it would contaminate our interpretation of the height effect and violate the exclusion restriction in the IV models.  This is a difficult problem to address, precisely because income and risk aversion move inversely, and traditional measures of risk-taking behavior will therefore move with income.  Complicating any attempt to directly assess this channel, the BHPS does not attempt to capture risk aversion parameters (such as relative risk tolerance, elicited in the Health and Retirement Study).  The closest approximation is a question from wave 18 which asks respondents “Are you generally a person who is fully prepared to take risks or do you try to avoid taking risks?”  This does not capture a standard economic parameter because it is endogenous to the types of risky choices that are available to the respondent.  Thus, someone who faces or is only familiar with poor risks (such as playing the lottery) may respond in a way that indicates risk aversion, whereas that same individual when offered the opportunity to take more sound risky decisions (such as investing in mutual funds versus a CD) might respond in the affirmative.   

With these caveats, in Table 3 below we produce models of Conservative support (corresponding to Tables 1 and 5 in the paper) which also include self-reported risk tolerance (using this variable) as a control variable.  The BHPS question is asked so that the respondent can answer anywhere between 1 (“unwilling to take risks”) and 10 (“fully prepared to take risks”).  We find that 5 is both the median and modal response, motivating us to code risk tolerance as a dummy in which an answer from 6-10 on the BHPS question represents a willingness to take risks.  We then employ this dummy as a control variable.   Results are reported in Table 3.  The inclusion of risk tolerance as a control does not alter the coefficient on height seen in the main paper (see Column 2 of Table 1 in the main paper).  Similarly, when we control for risk tolerance, the coefficient on income instrumented with height increases slightly (compare column 2 of Table H.3 here with column 2 of Table 5 in the main paper).  

\begin{sidewaystable}[ht]
\begin{scriptsize}
\begin{center}
\input{height_riskdummy.tex}
	\label{tab:risk}
\caption{Reduced Form: Height and Risk Tolerance}
\end{center}
Note: Dependent variable is \defcvote.  Column (1) reports least squares regression of ``supports Conservative party'' on height, controlling for risk tolerance. Column (2) reports two-stage least squares regression of ``supports Conservative party'' on income instrumented with height, also controlling for risk-tolerance.  All specifications include region dummies and heteroskedasticity-robust standard errors, clustered by household. \statsig.
\end{scriptsize}
\end{sidewaystable}






\newpage

\section{Notes on empirical analysis}
All empirical analysis was performed using Stata MP 12.1.  
\begin{itemize}
	\item Instrumental variables regressions are produced with \url{ivreg2} \citep{ivreg2}, with Anderson-Rubin confidence intervals constructed using \url{rivtest} \citep{rivtest_sj}.
	\item Running line smooths are generated using \url{mrunning} \citep{mrunning_sj}.  Pointwise confidence intervals are produced by 1000 replication bootstrap. 
	\item Tests of orthogonality conditions required by random effects estimators are implemented using \url{xtoverid} \citep{xtoverid}.
  \item Tables are produced using \url{esttab} \citep{esttab} and \url{est2tex} \citep{est2tex}.
\end{itemize}

\bibliography{federalism}


\end{document}